[lm-sensors] serious w83627ehf problem stopping cpu fan

[lm-sensors] serious w83627ehf problem stopping cpu fan

[Helmut Grohne]

Hi,

(I'm posting from outside, so please CC me on a reply.)

Some time ago I updated my kernel to 2.6.23.14 (vanilla) from something
pre-20. My system has a w83627ehf chip (the chip is labeled like that)
and the support for this chip seemed to have improved.

However my cpu fan "occasionally" stops spinning. This happened three
times at the time of this writing. It seems to happen when the system is
completely idle. As a result the cpu heats up and maybe destroying or
damaging itself over time. This is BAD!

I therefore tried to play with the /sys/devices/platform/w83627ehf.*/
interface. Changing the files pwm2* "sometimes" has an barely
predictable effect on the cpu fan. I was for instance able to start the
fan again by writing 2 to pwm2_enable the last time it stopped. However
I cannot disable the fan now by writing 0 to pwm2 which sometimes worked
before. Most of my writing data to these files seems to get ignored.

The only thing I really need now is a stable cpu fan. Can I somehow lock
it to maximum speed or something similar? Can I somehow run w83627ehf
with reduced functionality (like no longer modifying the fan at all)?

Please help me debug this. Do you want any other information not found
below?

Helmut Grohne


root@# i2cdetect
...
  Installed I2C busses:
    i2c-0       smbus           SMBus Via Pro adapter at 5000
root@# i2cdetect 0
WARNING! This program can confuse your I2C bus, cause data loss and worse!
I will probe file /dev/i2c-0.
I will probe address range 0x03-0x77.
Continue? [Y/n] y
     0  1  2  3  4  5  6  7  8  9  a  b  c  d  e  f
00:          -- -- -- -- -- -- -- -- -- -- -- -- -- 
10: -- -- -- -- -- -- -- -- -- -- -- -- -- -- -- -- 
20: -- -- -- -- -- -- -- -- -- -- -- -- -- -- -- -- 
30: -- -- -- -- -- -- -- -- -- -- -- -- -- -- -- -- 
40: -- -- -- -- -- -- -- -- -- -- -- -- -- -- -- -- 
50: 50 51 -- -- -- -- -- -- -- -- -- -- -- -- -- -- 
60: -- -- -- -- -- -- -- -- -- 69 -- -- -- -- -- -- 
70: -- -- -- -- -- -- -- --                         
root@# isadump 0x295 0x296
WARNING! Running this program can cause system crashes, data loss and worse!
I will probe address register 0x295 and data register 0x296.
Continue? [Y/n] y
     0  1  2  3  4  5  6  7  8  9  a  b  c  d  e  f
00: 01 01 01 ff 03 00 01 30 01 b1 01 01 3c 3c 01 01 
10: 01 ff 10 01 00 01 01 3c 43 07 00 00 57 00 00 c7 
20: 98 e4 cf cf ff c7 c7 26 ff 04 ff 01 00 01 df fd 
30: ef 5f b6 01 dc 01 01 01 ed 01 28 01 05 01 1f ff 
40: 01 ff a8 de ff ff 00 f0 2d ff 40 c1 10 15 01 a3 
50: 2c 00 00 01 00 50 00 ff ff ff ff ff ff ff ff ff 
60: 01 ff 40 01 01 01 3c ff 01 ff 01 ff ff ff ff ff 
70: ff ff ff ff ff ff ff ff ff ff ff ff ff ff ff ff 
root@# cat sensor_fix.sh 
#!/bin/bash
# this script is now invoked every 5 minutes from cron to try and fix
# the fan automatically.

cd /sys/devices/platform/w83627ehf.* || exit 1

test $(<fan2_input) -le 2000 || exit 0

echo fan not spinning

while :; do
        echo "trying pwm2_enable 2"
        echo 2 > pwm2_enable
        sleep 1
        test $(<fan2_input) -le 2000 || exit 0

        echo "trying pwm2 255"
        echo 255 > pwm2
        sleep 1
        test $(<fan2_input) -le 2000 || exit 0

        echo "trying pwm2_enable 1"
        echo 1 > pwm2_enable
        sleep 1
        test $(<fan2_input) -le 2000 || exit 0

        echo "trying pwm2 255"
        echo 255 > pwm2
        sleep 1
        test $(<fan2_input) -le 2000 || exit 0

        echo "trying pwm2_target 1000"
        echo 1000 > pwm2_target
        sleep 1
        test $(<fan2_input) -le 2000 || exit 0
done
root@# exit
helmut@$ dmesg
...
w83627ehf: Found W83627EHG chip at 0x290
...
helmut@$ dpkg -k lm-sensors
...
ii  lm-sensors 1:3.0.1-2 utilities to read temperature/voltage/fan sensors
helmut@$ lspci
00:00.0 Host bridge: VIA Technologies, Inc. K8T890 Host Bridge
00:00.1 Host bridge: VIA Technologies, Inc. K8T890 Host Bridge
00:00.2 Host bridge: VIA Technologies, Inc. K8T890 Host Bridge
00:00.3 Host bridge: VIA Technologies, Inc. K8T890 Host Bridge
00:00.4 Host bridge: VIA Technologies, Inc. K8T890 Host Bridge
00:00.5 PIC: VIA Technologies, Inc. K8T890 I/O APIC Interrupt Controller
00:00.7 Host bridge: VIA Technologies, Inc. K8T890 Host Bridge
00:01.0 PCI bridge: VIA Technologies, Inc. VT8237 PCI bridge [K8T800/K8T890 South]
00:02.0 PCI bridge: VIA Technologies, Inc. K8T890 PCI to PCI Bridge Controller
00:03.0 PCI bridge: VIA Technologies, Inc. K8T890 PCI to PCI Bridge Controller
00:03.1 PCI bridge: VIA Technologies, Inc. K8T890 PCI to PCI Bridge Controller
00:03.2 PCI bridge: VIA Technologies, Inc. K8T890 PCI to PCI Bridge Controller
00:03.3 PCI bridge: VIA Technologies, Inc. K8T890 PCI to PCI Bridge Controller
00:0c.0 Ethernet controller: Realtek Semiconductor Co., Ltd. RTL-8139/8139C/8139C+ (rev 10)
00:0f.0 IDE interface: VIA Technologies, Inc. VIA VT6420 SATA RAID Controller (rev 80)
00:0f.1 IDE interface: VIA Technologies, Inc. VT82C586A/B/VT82C686/A/B/VT823x/A/C PIPC Bus Master IDE (rev 06)
00:10.0 USB Controller: VIA Technologies, Inc. VT82xxxxx UHCI USB 1.1 Controller (rev 81)
00:10.1 USB Controller: VIA Technologies, Inc. VT82xxxxx UHCI USB 1.1 Controller (rev 81)
00:10.2 USB Controller: VIA Technologies, Inc. VT82xxxxx UHCI USB 1.1 Controller (rev 81)
00:10.3 USB Controller: VIA Technologies, Inc. VT82xxxxx UHCI USB 1.1 Controller (rev 81)
00:10.4 USB Controller: VIA Technologies, Inc. USB 2.0 (rev 86)
00:11.0 ISA bridge: VIA Technologies, Inc. VT8237 ISA bridge [KT600/K8T800/K8T890 South]
00:11.5 Multimedia audio controller: VIA Technologies, Inc. VT8233/A/8235/8237 AC97 Audio Controller (rev 60)
00:11.6 Communication controller: VIA Technologies, Inc. AC'97 Modem Controller (rev 80)
00:18.0 Host bridge: Advanced Micro Devices [AMD] K8 [Athlon64/Opteron] HyperTransport Technology Configuration
00:18.1 Host bridge: Advanced Micro Devices [AMD] K8 [Athlon64/Opteron] Address Map
00:18.2 Host bridge: Advanced Micro Devices [AMD] K8 [Athlon64/Opteron] DRAM Controller
00:18.3 Host bridge: Advanced Micro Devices [AMD] K8 [Athlon64/Opteron] Miscellaneous Control
02:00.0 VGA compatible controller: nVidia Corporation NV44 [GeForce 6200 TurboCache(TM)] (rev a1)
05:00.0 Ethernet controller: Marvell Technology Group Ltd. 88E8053 PCI-E Gigabit Ethernet Controller (rev 19)
helmut@$ lsmod
...
w83627ehf              26184  0
ac                      6728  0 
cpufreq_userspace       5228  0
cpufreq_ondemand       10064  1
cpufreq_powersave       2368  0
powernow_k8            15584  1
freq_table              5520  3 cpufreq_stats,cpufreq_ondemand,powernow_k8
hwmon_vid               4160  1 w83627ehf
i2c_viapro             10584  0 
k8temp                  6848  0
hwmon                   4104  2 w83627ehf,k8temp
i2c_core               28800  1 i2c_viapro
thermal                18336  0 
processor              38952  2 powernow_k8,thermal
fan                     6216  0
...
helmut@$

_______________________________________________
lm-sensors mailing list
lm-sensors@lm-sensors.org
http://lists.lm-sensors.org/mailman/listinfo/lm-sensors

Re: [lm-sensors] serious w83627ehf problem stopping cpu fan

[Rudolf Marek]

Hi,

What motherboard it is? Asus A8V-E SE? Can you please send dsdt table here?

cd /tmp
cat /proc/acpi/dsdt  >dsdt.bin
iasl -d dsdt.bin

(output will be named dsdt.dsl)

If you go to BIOS, is Q-fan enabled? What is your settings there?

Thanks,
Rudolf


_______________________________________________
lm-sensors mailing list
lm-sensors@lm-sensors.org
http://lists.lm-sensors.org/mailman/listinfo/lm-sensors

Re: [lm-sensors] serious w83627ehf problem stopping cpu fan

[Helmut Grohne]

Hi,

thanks for your reply and your help!

> What motherboard it is? Asus A8V-E SE? Can you please send dsdt table here?

The mainboard is labeled A8V-E and the box it came in is labeled A8V-E SE.

> cd /tmp
> cat /proc/acpi/dsdt  >dsdt.bin
> iasl -d dsdt.bin

Unfortunately there is no /proc/acpi/dsdt on my system. Can you tell me
what is missing? A kernel option?

> If you go to BIOS, is Q-fan enabled? What is your settings there?

Yes I know that Q-fan is enabled. A (more hardware savvy) friend of mine
configured this. The target temperature is 40°C with a tolerance of 3°C.
A setting I do not understand is called "fan duty cycle" and is set to
11/16.

My fan used to spin between 3000 and 5000 rpm. However the system
somewhen[1] stopped doing so and now merely sits on about 3500 however
hot the cpu is. This is also bad, because I cannot run the system at
full speed with work for a longer time.

Helmut

[1] Kernels I used: 2.6.18, since March 2007 2.6.20.1, since January
    2.6.23.14. According to munin graphs the system stopped spinning
    higher than 4000 rpm around September.

_______________________________________________
lm-sensors mailing list
lm-sensors@lm-sensors.org
http://lists.lm-sensors.org/mailman/listinfo/lm-sensors

Re: [lm-sensors] serious w83627ehf problem stopping cpu fan

[Jean Delvare]

Hallo Helmut,

On Wed, 12 Mar 2008 11:52:18 +0100, Helmut Grohne wrote:
> Unfortunately there is no /proc/acpi/dsdt on my system. Can you tell me
> what is missing? A kernel option?

Recent kernel export it as /sys/firmware/acpi/tables/DSDT instead.

-- 
Jean Delvare

_______________________________________________
lm-sensors mailing list
lm-sensors@lm-sensors.org
http://lists.lm-sensors.org/mailman/listinfo/lm-sensors

Re: [lm-sensors] serious w83627ehf problem stopping cpu fan

[Helmut Grohne]

Hi,

> cd /tmp
> cat /proc/acpi/dsdt  >dsdt.bin
> iasl -d dsdt.bin

Thanks to Jean Delvare for pointing me to
/sys/firmware/acpi/tables/DSDT. This now worked.


> (output will be named dsdt.dsl)

The file can be found at http://subdivi.de/~helmut/dsdt.dsl.

Helmut

_______________________________________________
lm-sensors mailing list
lm-sensors@lm-sensors.org
http://lists.lm-sensors.org/mailman/listinfo/lm-sensors

Re: [lm-sensors] serious w83627ehf problem stopping cpu fan

[Jean Delvare]

On Wed, 12 Mar 2008 02:14:32 +0100, Helmut Grohne wrote:
> Hi,
> 
> (I'm posting from outside, so please CC me on a reply.)
> 
> Some time ago I updated my kernel to 2.6.23.14 (vanilla) from something
> pre-20. My system has a w83627ehf chip (the chip is labeled like that)
> and the support for this chip seemed to have improved.
> 
> However my cpu fan "occasionally" stops spinning. This happened three
> times at the time of this writing. It seems to happen when the system is
> completely idle. As a result the cpu heats up and maybe destroying or
> damaging itself over time. This is BAD!
> 
> I therefore tried to play with the /sys/devices/platform/w83627ehf.*/
> interface. Changing the files pwm2* "sometimes" has an barely
> predictable effect on the cpu fan. I was for instance able to start the
> fan again by writing 2 to pwm2_enable the last time it stopped. However
> I cannot disable the fan now by writing 0 to pwm2 which sometimes worked
> before. Most of my writing data to these files seems to get ignored.
> 
> The only thing I really need now is a stable cpu fan. Can I somehow lock
> it to maximum speed or something similar? Can I somehow run w83627ehf

If you disable Q-Fan in the BIOS, that's what should happen.

> with reduced functionality (like no longer modifying the fan at all)?

No. The w83627ehf driver doesn't change the chip settings by itself, so
there's nothing to reduce.

> 
> Please help me debug this. Do you want any other information not found
> below?

Did you try without the w83627ehf driver loaded? What makes you believe
that the w83627ehf driver is responsible for your problem? This might
as well be an ACPI issue.

> 
> Helmut Grohne
> 
> (...)
> root@# cat sensor_fix.sh 
> #!/bin/bash
> # this script is now invoked every 5 minutes from cron to try and fix
> # the fan automatically.
> 
> cd /sys/devices/platform/w83627ehf.* || exit 1
> 
> test $(<fan2_input) -le 2000 || exit 0
> 
> echo fan not spinning
> 
> while :; do
>         echo "trying pwm2_enable 2"
>         echo 2 > pwm2_enable
>         sleep 1
>         test $(<fan2_input) -le 2000 || exit 0
> 
>         echo "trying pwm2 255"
>         echo 255 > pwm2
>         sleep 1
>         test $(<fan2_input) -le 2000 || exit 0

Note that the value written to pwm2 has no effect when pwm2_enable = 2,
so there's nothing you can hope from this step.

> 
>         echo "trying pwm2_enable 1"
>         echo 1 > pwm2_enable
>         sleep 1
>         test $(<fan2_input) -le 2000 || exit 0
> 
>         echo "trying pwm2 255"
>         echo 255 > pwm2
>         sleep 1
>         test $(<fan2_input) -le 2000 || exit 0
> 
>         echo "trying pwm2_target 1000"
>         echo 1000 > pwm2_target
>         sleep 1
>         test $(<fan2_input) -le 2000 || exit 0

This is setting the target temperature to 1 degree C. That doesn't make
any sense to me, as there is the way the chip can achieve this. On top
of that, this setting is only valid fir pwm2_enable = 2, which isn't
the case at this point.

> done
> root@# exit

-- 
Jean Delvare

_______________________________________________
lm-sensors mailing list
lm-sensors@lm-sensors.org
http://lists.lm-sensors.org/mailman/listinfo/lm-sensors

Re: [lm-sensors] serious w83627ehf problem stopping cpu fan

[Helmut Grohne]

Hi,

> > The only thing I really need now is a stable cpu fan. Can I somehow lock
> > it to maximum speed or something similar? Can I somehow run w83627ehf
> If you disable Q-Fan in the BIOS, that's what should happen.

Ok.

> Did you try without the w83627ehf driver loaded? What makes you believe
> that the w83627ehf driver is responsible for your problem? This might
> as well be an ACPI issue.

I did not try unloading the driver permanently, because this would
effectively prevent me from monitoring temperature and fan speed and
would therefore not notice a temperature increase.

There are two things that made me believe w83627ehf is responsible:
1) It is possible to change the fan behaviour using this driver.
2) This is very unreliable. (see below)

I did not think about this having to do with ACPI, but it sounds
reasonable. Compiling dsdt.dsl again results in tons of errors (below).

> >         echo 255 > pwm2
> Note that the value written to pwm2 has no effect when pwm2_enable = 2,
> so there's nothing you can hope from this step.

Even with pwm2_enable = 1 the setting often does not change
(nondeterministically). That's the reason for doing these echos in a
loop: Hoping that a value will eventually change.

> >         echo 1000 > pwm2_target
> This is setting the target temperature to 1 degree C. That doesn't make
> any sense to me, as there is the way the chip can achieve this. On top
> of that, this setting is only valid fir pwm2_enable = 2, which isn't
> the case at this point.

What makes even less sense is taht pwm2_target only accepts two values:
1000 and 127000. Other values I tried simply do not change the contents
of the file.

Theoretically the fan should stop after the following.
echo 1 > pwm2_enable
echo 0 > pwm2

This works only in very rare occasions. Values written to pwm2 seem to
only sometimes change the value.

Helmut

root@# iasl -tc dsdt.dsl
Intel ACPI Component Architecture
ASL Optimizing Compiler version 20061109 [May 15 2007]
Copyright (C) 2000 - 2006 Intel Corporation
Supports ACPI Specification Revision 3.0a

dsdt.dsl   410:     Method (\_WAK, 1, NotSerialized)
Warning  1079 -                 ^ Reserved method must return a value (_WAK)

dsdt.dsl   452:             Store (Local0, Local0)
Error    4049 -                         ^ Method local variable is not initialized (Local0)

dsdt.dsl   460:             Store (Local0, Local0)
Error    4049 -                         ^ Method local variable is not initialized (Local0)

dsdt.dsl   786:                 Method (GHTF, 0, NotSerialized)
Warning  1086 -                            ^ Not all control paths return a value (GHTF)

dsdt.dsl  1697:                         Store (Local0, Local0)
Error    4049 -                                     ^ Method local variable is not initialized (Local0)

dsdt.dsl  1702:                         Store (Local0, Local0)
Error    4049 -                                     ^ Method local variable is not initialized (Local0)

dsdt.dsl  1707:                         Store (Local0, Local0)
Error    4049 -                                     ^ Method local variable is not initialized (Local0)

dsdt.dsl  2198:                 Method (STM, 0, Serialized)
Warning  1086 -                           ^ Not all control paths return a value (STM_)

dsdt.dsl  2757:                 Method (STM, 0, Serialized)
Warning  1086 -                           ^ Not all control paths return a value (STM_)

dsdt.dsl  3864:                 Method (RVLT, 1, NotSerialized)
Warning  1086 -                            ^ Not all control paths return a value (RVLT)

dsdt.dsl  3934:                 Method (RTMP, 1, NotSerialized)
Warning  1086 -                            ^ Not all control paths return a value (RTMP)

dsdt.dsl  4111:                     Store (GHTF (), Local1)
Warning  1091 -                               ^ Called method may not always return a value

dsdt.dsl  4156:                 Method (OCOP, 1, NotSerialized)
Warning  1086 -                            ^ Not all control paths return a value (OCOP)

dsdt.dsl  4182:                             Store (GHTF (), Local1)
Warning  1091 -                                       ^ Called method may not always return a value

dsdt.dsl  5963:                 Method (_STA, 0, NotSerialized)
Warning  1086 -                            ^ Not all control paths return a value (_STA)

dsdt.dsl  5963:                 Method (_STA, 0, NotSerialized)
Warning  1079 -                            ^ Reserved method must return a value (_STA)

dsdt.dsl  6035:                 Method (_PRS, 0, NotSerialized)
Warning  1086 -                            ^ Not all control paths return a value (_PRS)

dsdt.dsl  6035:                 Method (_PRS, 0, NotSerialized)
Warning  1079 -                            ^ Reserved method must return a value (_PRS)

dsdt.dsl  6111:                 Method (_STA, 0, NotSerialized)
Warning  1086 -                            ^ Not all control paths return a value (_STA)

dsdt.dsl  6111:                 Method (_STA, 0, NotSerialized)
Warning  1079 -                            ^ Reserved method must return a value (_STA)

dsdt.dsl  6166:                 Method (_PRS, 0, NotSerialized)
Warning  1086 -                            ^ Not all control paths return a value (_PRS)

dsdt.dsl  6166:                 Method (_PRS, 0, NotSerialized)
Warning  1079 -                            ^ Reserved method must return a value (_PRS)

dsdt.dsl  6293:                 Method (_PRS, 0, NotSerialized)
Warning  1086 -                            ^ Not all control paths return a value (_PRS)

dsdt.dsl  6293:                 Method (_PRS, 0, NotSerialized)
Warning  1079 -                            ^ Reserved method must return a value (_PRS)

dsdt.dsl  6429:                 Method (_PRS, 0, NotSerialized)
Warning  1086 -                            ^ Not all control paths return a value (_PRS)

dsdt.dsl  6429:                 Method (_PRS, 0, NotSerialized)
Warning  1079 -                            ^ Reserved method must return a value (_PRS)

dsdt.dsl  6475:                 Method (_CRS, 0, NotSerialized)
Warning  1086 -                            ^ Not all control paths return a value (_CRS)

dsdt.dsl  6475:                 Method (_CRS, 0, NotSerialized)
Warning  1079 -                            ^ Reserved method must return a value (_CRS)

ASL Input:  dsdt.dsl - 8614 lines, 278641 bytes, 3542 keywords
Compilation complete. 5 Errors, 23 Warnings, 0 Remarks, 1099 Optimizations
root@#

_______________________________________________
lm-sensors mailing list
lm-sensors@lm-sensors.org
http://lists.lm-sensors.org/mailman/listinfo/lm-sensors

Re: [lm-sensors] serious w83627ehf problem stopping cpu fan

[Rudolf Marek]

Hi,

Just load k8temp driver, it will tell you the CPU temperature. The CPU will shut 
down when too hot - the shutdown point is around 130C

Rudolf

_______________________________________________
lm-sensors mailing list
lm-sensors@lm-sensors.org
http://lists.lm-sensors.org/mailman/listinfo/lm-sensors

Re: [lm-sensors] serious w83627ehf problem stopping cpu fan

[Helmut Grohne]

Hi,

This is a summary of the irc discussion on #linux-sensors with Jean
Delvare and Rudolf Ruik Marek.

First and most important: Unloading the thermal module makes the /sys
interface of w83627ehf behave normally, so the problem boils down to
the module interfering with each other.

Secondly my dsdt is buggy. I do not know whether the bugs are visible.
Errors like this one look like they may have an effect:

dsdt.dsl   786:                 Method (GHTF, 0, NotSerialized)
Warning  1086 -                            ^ Not all control paths return a value (GHTF)

Helmut

_______________________________________________
lm-sensors mailing list
lm-sensors@lm-sensors.org
http://lists.lm-sensors.org/mailman/listinfo/lm-sensors